S. C. Warda, M. B. Hursthousea, D. C. Woodsb and S. M. Lewisb a School of Chemistry, University of Southampton, Southampton, SO17 1BJ, UK.

b School of Mathematics, University of Southampton, Southampton, SO17 1BJ, UK.

There is currently a great amount of interest in the use of salts in the pharmaceutical industry because certain properties of the solid forms can be modified without altering the desired effect of the drug[1]. Salt formation is essentially a three component system involving an acid (A), a base (B) and one or more solvents. A salt is formed by the transfer of a proton (H+) from an acid (A) to a base (B): A-H + B → (A-)(B+-H) . In order to assist salt selection a number of empirical rules have been devised, such as the ‘rule of three’. This states that salt formation generally requires a difference of at least three pK a units between the conjugate base and the conjugate acid, pKa (Base) - pKa (Acid) ≥ 3, where pKa is the ability of an ionisable group to donate a proton (H+) in an aqueous medium and is often referred to as the dissociation constant. Although such rules are valuable guidelines, as far as we are aware no detailed study has been made of the reliability and/or basis of these empirical procedures. A carefully planned set of experiments may lead to a more scientific method for assessing the viability of salt formation, rather than relying on trial and error.

INVESTIGATING THE CSD

The Cambridge Structural Database (CSD)[2] was searched to identify acids which form salts and co-crystals. The objective was to identify when a salt forms in preference to a co-crystal. As an example, we consider succinic acid (Figure 1).

2 polymorphs5 organic salts 8 co-crystals

NH NH

Salt forming base

Diagram pK1

piperazine 9.8 B

Doxylamine 8.7 B

L-Lysine 9.5 B

Imidazole(Figure 2)

6.9 B

Ethylene-1,2-diammonium

9.9 B

NO

NCH3

CH3CH3

Ph

NH3+

O

NH2

O

NNH

NH2 NH2

Co-crystal forming base

Diagram pK1

Benzamide *

2-pyridone 0.7 B

picolinylaminocyclohex-2-enone

5.7 B

aminopyrimidine(Figure 3)

3.9 B

cyanophenylimidazolylmethane

6.1 B

Urea 0.1 B

Phenazine 1.6 B

2-amino-6-ethyl-4(3H)-pyrpyrimidone

0.5 B

NH2

O

Ph

NH

O

N

O

NH

N

N

NH2

NN

NC

CN

O

N

N

N

N

NH

N

O

NH2 Et

Tables 1 and 2. Salt and co-crystal forming bases for succinic acid with their pK1[4] values, the pKa for the most ionisable site in the molecule. B: indicates a basic pKa. * pK1 of this compound has not been reported in the literature.

DESCRIPTORS

We are investigating an approach in which a statistical model, called a response surface model[7], is fitted to the data from a designed experiment. The fitted model may then be used to predict combinations of acids and bases that are likely to produce a salt. Due to the fact that there is a wide variety of choices for the acid or the base, a set of chemical descriptors was sought that could be used to characterise the chemical space of interest and to form the statistical model. The chosen descriptors should represent key aspects of the molecular structure, which relate to its salt forming ability. A shortlist of such descriptors was eventually chosen that were tabulated in the literature or easily calculated.

CONCLUSION

A set of descriptors for investigating salt formation has been identified. These descriptors can now be used in experiments to investigate the properties needed for salt formation to occur. The eventual aim is that a database containing the descriptor values will be available to the scientific community over the Grid, together with rules that indicate which counter-ion would be most appropriate for a drug with a specified set of descriptors. It is also planned to make available search algorithms for finding suitable designs for the experiments via a software node on the Grid.

ACKNOWLEDGEMENTS

We gratefully acknowledge the support of the EPSRC e-Science programme (GR/R67729, Combechem) along with AstraZeneca in Mölndal, Sweden.

1. S. M. Berge, L. D. Bighley, and D. C. Monkhouse, J. Pharm. Sci., 1977, 66, 1.

2. F. H. Allen, Acta Crystallogr. Sect. B, 2002, 58, 380-388.3. N. N. Petropavlov, S. B. Yarantsev, Kristallografiya, 1983,

28, 1132.4. D. D. Perrin, Dissociation Constants of Organic Bases in

Aqueous Solution: Supplement. 1972, London: Butterworths Scientific Publications.

Succinic acid has a pK1 value of 4.2 and can exist in two polymorphic forms[3]; a total of five salts and eight co-crystals were found from the CSD, using specific criteria (Tables 1 and 2). A co-crystal is taken to be an A-B composite in which no proton transfer has occurred:

A-H + B → (A-H)(B)

Figure 2. A packing diagram for imidazolium hydrogen succinate viewed along the c axis[5].

Figure 3. A packing diagram for 2-aminopyrimidine succinic acid, viewed along the a axis[6].

As a starting point, an initial set of 67 acids was obtained using the CSD. Values for the selected descriptors were either found in the literature or calculated using software such as HyperChem[8]. Values for a total of ten descriptors were investigated. Figure 4 shows a matrix of plots of all the two-dimensional projections (scatter plots) of the values of the ten descriptors (labelled X1 to X10 for simplicity) for the acids.

These scatter plots show the relationship between pairs of descriptors for the available acids. A high proportion of points along the diagonal indicates a strong correlation between two descriptors. Descriptors were then removed from the model to avoid any strong relationships, such as highly correlated pairs. This resulted in descriptors X1, X2, X3, X5, together with either X8 or X9. The next step in the process was to extend the set of acids to obtain better coverage of salt formation space. This was achieved by first identifying regions in the descriptor space where acids were sparse and then finding additional acids in these regions.

Figure 4. Two-dimensional projections for X1 – X10.

EXPERIMENTAL DESIGN

For our initial set of 67 compounds, and using the descriptors X1, X2, X3, X5 and X8, Figure 5 shows the two-dimensional projections of a 24 point coverage design. The points in the coverage design are chosen to ensure that each unselected compound is as close to a selected compound as possible, giving similar, but less dense, projections compared with Figure 4.

The example given using succinic acid agrees with the hypothesis that a large difference in pKa values leads to a higher chance of salt formation.

Figure 1. Succinic acid.

Figure 5. Two-dimensional projections for a coverage design for five descriptors.

5. J. C. MacDonald, P. C. Dorrestein and M.M. Pilley, Cryst. Growth Des., 2001, 1, 29-38.

6. M. C. Etter, D. A. Admond and Doyle Britton, Acta Cryst., 1990, C46, 933-934.

7. R. H. Myers and D. C. Montgomery, Response Surface Methodology (2nd ed.), 2002, New York:Wiley.

8. HyperChem, Inc. 115 NW 4th Street, Gainsville, Florida 32601, USA.

REFERENCES